Purnima Kamath's Avatar

Purnima Kamath

@purnimakamath

Postdoc @ NYU MARL | Ph.Done. @ NUS purnimakamath.com

34
Followers
129
Following
7
Posts
22.10.2024
Joined
Posts Following

Latest posts by Purnima Kamath @purnimakamath

If you thought arrival, immigration, and exit at Changi Airport was fast - Voting in Singapore is even faster! In and out of the polling station in record 5 mins this morning.
#HappyElectionDay

03.05.2025 07:03 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0

A better way of expressing why LLMs arent generating new knowledge : "They're (LLMs) mostly doing "manifold filling" at the moment - filling in the interpolation gaps between what humans already know, somehow treating knowledge as an intangible fabric of reality."
thomwolf.io/blog/scienti...

25.03.2025 01:38 ๐Ÿ‘ 1 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0

This is true if only assuming that the "un-symbolized" or unrepresented excess is in-between our existing symbols or words.

10.01.2025 03:53 ๐Ÿ‘ 0 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 1

Makes one wonder how we decided to build an AGI using language as the philosophers (Derrida, Lacan) of our past have been saying that language is insufficient to represent reality.

10.01.2025 01:19 ๐Ÿ‘ 2 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 1

A Jacques Lacan-esque research orientation with the idea that language is not enough. Words are insufficient for communication and our thoughts are always guided by an excess - something that cannot be put into words.

10.01.2025 01:19 ๐Ÿ‘ 0 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0

An alternate history (fiction) I will be interested in reading is one where chatgpt didn't happen. An inflection point in our AI research lives where every discourse does not surround language or language-based models or agents.

10.01.2025 01:19 ๐Ÿ‘ 0 ๐Ÿ” 0 ๐Ÿ’ฌ 1 ๐Ÿ“Œ 0
Preview
MorphFader: Enabling Fine-grained Controllable Morphing with Text-to-Audio Models Sound morphing is the process of gradually and smoothly transforming one sound into another to generate novel and perceptually hybrid sounds that simultaneously resemble both. Recently, diffusion-base...

Stoked to report that our paper "MorphFader: Enabling Fine-grained Controllable Morphing with Text-to-Audio Models" has been accepted at ICASSP 2025! My final paper from my PhD thesis at NUS!

arxiv.org/abs/2408.07260

See you all at ICASSP in Hyderabad, India in April 2025! #icassp2025

21.12.2024 04:14 ๐Ÿ‘ 3 ๐Ÿ” 0 ๐Ÿ’ฌ 0 ๐Ÿ“Œ 0